Cleanup microcluster/k8s when scaling down cluster nodes

bootstrap/controllers/ck8sconfig_controller.go

@@ -220,22 +220,12 @@ func (r *CK8sConfigReconciler) joinControlplane(ctx context.Context, scope *Scop
 	// injects into config.Version values from top level object
 	r.reconcileTopLevelObjectSettings(scope.Cluster, machine, scope.Config)
 
-	authToken, err := token.Lookup(ctx, r.Client, client.ObjectKeyFromObject(scope.Cluster))
-	if err != nil {
-		conditions.MarkFalse(scope.Config, bootstrapv1.DataSecretAvailableCondition, bootstrapv1.DataSecretGenerationFailedReason, clusterv1.ConditionSeverityWarning, err.Error())
-		return err
-	}
-
-	if authToken == nil {
-		return fmt.Errorf("auth token not yet generated")
-	}
-
-	workloadCluster, err := r.managementCluster.GetWorkloadCluster(ctx, util.ObjectKey(scope.Cluster))
+	workloadCluster, err := r.managementCluster.GetWorkloadCluster(ctx, util.ObjectKey(scope.Cluster), scope.Config.Spec.ControlPlaneConfig.MicroclusterPort)
 	if err != nil {
 		return fmt.Errorf("failed to create remote cluster client: %w", err)
 	}
 
-	joinToken, err := workloadCluster.NewControlPlaneJoinToken(ctx, *authToken, scope.Config.Spec.ControlPlaneConfig.MicroclusterPort, scope.Config.Name)
+	joinToken, err := workloadCluster.NewControlPlaneJoinToken(ctx, scope.Config.Name)
 	if err != nil {
 		return fmt.Errorf("failed to request join token: %w", err)
 	}
@@ -303,12 +293,12 @@ func (r *CK8sConfigReconciler) joinWorker(ctx context.Context, scope *Scope) err
 		return fmt.Errorf("auth token not yet generated")
 	}
 
-	workloadCluster, err := r.managementCluster.GetWorkloadCluster(ctx, util.ObjectKey(scope.Cluster))
+	workloadCluster, err := r.managementCluster.GetWorkloadCluster(ctx, util.ObjectKey(scope.Cluster), scope.Config.Spec.ControlPlaneConfig.MicroclusterPort)
 	if err != nil {
 		return fmt.Errorf("failed to create remote cluster client: %w", err)
 	}
 
-	joinToken, err := workloadCluster.NewWorkerJoinToken(ctx, *authToken, scope.Config.Spec.ControlPlaneConfig.MicroclusterPort, scope.Config.Name)
+	joinToken, err := workloadCluster.NewWorkerJoinToken(ctx)
 	if err != nil {
 		return fmt.Errorf("failed to request join token: %w", err)
 	}

controlplane/controllers/ck8scontrolplane_controller.go

@@ -299,7 +299,7 @@ func (r *CK8sControlPlaneReconciler) SetupWithManager(ctx context.Context, mgr c
 
 	if r.managementClusterUncached == nil {
 		r.managementClusterUncached = &ck8s.Management{
-			Client:          mgr.GetAPIReader(),
+			Client:          mgr.GetClient(),
 			K8sdDialTimeout: r.K8sdDialTimeout,
 		}
 	}
@@ -378,7 +378,8 @@ func (r *CK8sControlPlaneReconciler) updateStatus(ctx context.Context, kcp *cont
 		}
 	}
 
-	workloadCluster, err := r.managementCluster.GetWorkloadCluster(ctx, util.ObjectKey(cluster))
+	microclusterPort := kcp.Spec.CK8sConfigSpec.ControlPlaneConfig.MicroclusterPort
+	workloadCluster, err := r.managementCluster.GetWorkloadCluster(ctx, util.ObjectKey(cluster), microclusterPort)
 	if err != nil {
 		return fmt.Errorf("failed to create remote cluster client: %w", err)
 	}
@@ -665,7 +666,8 @@ func (r *CK8sControlPlaneReconciler) reconcileControlPlaneConditions(ctx context
 		return nil
 	}
 
-	workloadCluster, err := r.managementCluster.GetWorkloadCluster(ctx, util.ObjectKey(controlPlane.Cluster))
+	microclusterPort := controlPlane.KCP.Spec.CK8sConfigSpec.ControlPlaneConfig.MicroclusterPort
+	workloadCluster, err := r.managementCluster.GetWorkloadCluster(ctx, util.ObjectKey(controlPlane.Cluster), microclusterPort)
 	if err != nil {
 		return fmt.Errorf("cannot get remote client to workload cluster: %w", err)
 	}

controlplane/controllers/machine_controller.go

@@ -4,6 +4,7 @@ import (
 	"context"
 	"time"
 
+	"github.com/canonical/cluster-api-k8s/pkg/ck8s"
 	"github.com/go-logr/logr"
 	"github.com/pkg/errors"
 	apierrors "k8s.io/apimachinery/pkg/api/errors"
@@ -24,10 +25,7 @@ type MachineReconciler struct {
 
 	K8sdDialTimeout time.Duration
 
-	// NOTE(neoaggelos): See note below
-	/**
 	managementCluster ck8s.ManagementCluster
-	**/
 }
 
 func (r *MachineReconciler) SetupWithManager(ctx context.Context, mgr ctrl.Manager, log *logr.Logger) error {
@@ -36,15 +34,16 @@ func (r *MachineReconciler) SetupWithManager(ctx context.Context, mgr ctrl.Manag
 		Build(r)
 
 	// NOTE(neoaggelos): See note below
-	/**
 	if r.managementCluster == nil {
 		r.managementCluster = &ck8s.Management{
 			Client:          r.Client,
-			EtcdDialTimeout: r.EtcdDialTimeout,
-			EtcdCallTimeout: r.EtcdCallTimeout,
+			K8sdDialTimeout: r.K8sdDialTimeout,
+			/*
+				EtcdDialTimeout: r.EtcdDialTimeout,
+				EtcdCallTimeout: r.EtcdCallTimeout,
+			*/
 		}
 	}
-	**/
 
 	return err
 }
@@ -101,7 +100,6 @@ func (r *MachineReconciler) Reconcile(ctx context.Context, req ctrl.Request) (ct
 			logger.Info("unable to get cluster.")
 			return ctrl.Result{}, errors.Wrapf(err, "unable to get cluster")
 		}
-
 		workloadCluster, err := r.managementCluster.GetWorkloadCluster(ctx, util.ObjectKey(cluster))
 		if err != nil {
 			logger.Error(err, "failed to create client to workload cluster")

controlplane/controllers/remediation.go

@@ -28,6 +28,7 @@ import (
 	kerrors "k8s.io/apimachinery/pkg/util/errors"
 	"k8s.io/klog/v2"
 	clusterv1 "sigs.k8s.io/cluster-api/api/v1beta1"
+	"sigs.k8s.io/cluster-api/util"
 	"sigs.k8s.io/cluster-api/util/annotations"
 	"sigs.k8s.io/cluster-api/util/collections"
 	"sigs.k8s.io/cluster-api/util/conditions"
@@ -236,6 +237,18 @@ func (r *CK8sControlPlaneReconciler) reconcileUnhealthyMachines(ctx context.Cont
 		**/
 	}
 
+	microclusterPort := controlPlane.KCP.Spec.CK8sConfigSpec.ControlPlaneConfig.MicroclusterPort
+	clusterObjectKey := util.ObjectKey(controlPlane.Cluster)
+	workloadCluster, err := r.managementCluster.GetWorkloadCluster(ctx, clusterObjectKey, microclusterPort)
+	if err != nil {
+		log.Error(err, "failed to create client to workload cluster")
+		return ctrl.Result{}, errors.Wrapf(err, "failed to create client to workload cluster")
+	}
+
+	if err := workloadCluster.RemoveMachineFromCluster(ctx, machineToBeRemediated); err != nil {
+		log.Error(err, "failed to remove machine from microcluster")
+	}
+
 	// Delete the machine
 	if err := r.Client.Delete(ctx, machineToBeRemediated); err != nil {
 		conditions.MarkFalse(machineToBeRemediated, clusterv1.MachineOwnerRemediatedCondition, clusterv1.RemediationFailedReason, clusterv1.ConditionSeverityError, err.Error())

controlplane/controllers/scale.go

@@ -159,6 +159,18 @@ func (r *CK8sControlPlaneReconciler) scaleDownControlPlane(
 	}
 	**/
 
+	microclusterPort := controlPlane.KCP.Spec.CK8sConfigSpec.ControlPlaneConfig.MicroclusterPort
+	clusterObjectKey := util.ObjectKey(cluster)
+	workloadCluster, err := r.managementCluster.GetWorkloadCluster(ctx, clusterObjectKey, microclusterPort)
+	if err != nil {
+		logger.Error(err, "failed to create client to workload cluster")
+		return ctrl.Result{}, errors.Wrapf(err, "failed to create client to workload cluster")
+	}
+
+	if err := workloadCluster.RemoveMachineFromCluster(ctx, machineToDelete); err != nil {
+		logger.Error(err, "failed to remove machine from microcluster")
+	}
+
 	logger = logger.WithValues("machine", machineToDelete)
 	if err := r.Client.Delete(ctx, machineToDelete); err != nil && !apierrors.IsNotFound(err) {
 		logger.Error(err, "Failed to delete control plane machine")

controlplane/main.go

@@ -66,7 +66,7 @@ func main() {
 	flag.DurationVar(&syncPeriod, "sync-period", 10*time.Minute,
 		"The minimum interval at which watched resources are reconciled (e.g. 15m)")
 
-	flag.DurationVar(&k8sdDialTimeout, "k8sd-dial-timeout-duration", 10*time.Second,
+	flag.DurationVar(&k8sdDialTimeout, "k8sd-dial-timeout-duration", 60*time.Second,
 		"Duration that the proxy client waits at most to establish a connection with k8sd")
 
 	flag.Parse()

pkg/ck8s/api/cluster_node.go

@@ -0,0 +1,7 @@
+package apiv1
+
+// RemoveNodeRequest is used to request to remove a node from the cluster.
+type RemoveNodeRequest struct {
+	Name  string `json:"name"`
+	Force bool   `json:"force"`
+}

pkg/ck8s/management_cluster.go

@@ -5,6 +5,7 @@ import (
 	"fmt"
 	"time"
 
+	"github.com/canonical/cluster-api-k8s/pkg/token"
 	"k8s.io/client-go/kubernetes/scheme"
 	clusterv1 "sigs.k8s.io/cluster-api/api/v1beta1"
 	"sigs.k8s.io/cluster-api/controllers/remote"
@@ -17,14 +18,14 @@ type ManagementCluster interface {
 	client.Reader
 
 	GetMachinesForCluster(ctx context.Context, cluster client.ObjectKey, filters ...collections.Func) (collections.Machines, error)
-	GetWorkloadCluster(ctx context.Context, clusterKey client.ObjectKey) (*Workload, error)
+	GetWorkloadCluster(ctx context.Context, clusterKey client.ObjectKey, microclusterPort int) (*Workload, error)
 }
 
 // Management holds operations on the management cluster.
 type Management struct {
 	ManagementCluster
 
-	Client client.Reader
+	Client client.Client
 
 	K8sdDialTimeout time.Duration
 }
@@ -70,7 +71,7 @@ const (
 
 // GetWorkloadCluster builds a cluster object.
 // The cluster comes with an etcd client generator to connect to any etcd pod living on a managed machine.
-func (m *Management) GetWorkloadCluster(ctx context.Context, clusterKey client.ObjectKey) (*Workload, error) {
+func (m *Management) GetWorkloadCluster(ctx context.Context, clusterKey client.ObjectKey, microclusterPort int) (*Workload, error) {
 	restConfig, err := remote.RESTConfig(ctx, CK8sControlPlaneControllerName, m.Client, clusterKey)
 	if err != nil {
 		return nil, fmt.Errorf("failed to get config: %w", err)
@@ -87,10 +88,25 @@ func (m *Management) GetWorkloadCluster(ctx context.Context, clusterKey client.O
 		return nil, &RemoteClusterConnectionError{Name: clusterKey.String(), Err: err}
 	}
 
+	authToken, err := token.Lookup(ctx, m.Client, clusterKey)
+	if err != nil {
+		return nil, fmt.Errorf("failed to lookup auth token: %w", err)
+	}
+
+	if authToken == nil {
+		return nil, fmt.Errorf("auth token not yet generated")
+	}
+
+	if microclusterPort == 0 {
+		microclusterPort = 2380
+	}
+
 	workload := &Workload{
+		authToken:           *authToken,
 		Client:              c,
 		ClientRestConfig:    restConfig,
 		K8sdClientGenerator: g,
+		microclusterPort:    microclusterPort,
 
 		/**
 		CoreDNSMigrator: &CoreDNSMigrator{},
@@ -155,3 +171,5 @@ func (m *Management) getEtcdCAKeyPair(ctx context.Context, clusterKey client.Obj
 	return crtData, keyData, nil
 }
 **/
+
+var _ ManagementCluster = &Management{}